Shock Hunting: Effects of Regional-dependent, Regional-specific Shocks in the Swedish Property Market

Abstract

The purpose of this paper is to evaluate housing price risks by examining the price diffusion process in different regions of Sweden. We model regional housing volatility as stemming from two sources: regional-specific shocks that reflect fundamental changes and regional-dependent shocks that reflect shocks in related regions. We evaluate the relative contributions of these two sources of regional price variations and measure shock persistence using multivariate modelling. Our results confirm the importance of price diffusion across regions and indicate that several influential regional prices are singled out as powerful determinants of regional heterogeneous volatility through direct and indirect influences.

Key Words:

• Price risk
• Information diffusion
• Regional-specific shock
• Regional-dependent shock
• Shock persistence

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. This method has been applied to the analysis of output fluctuations and exchange rate shocks, as in Pesaran, Pierse, and Lee (1993), Van de Gucht, Dekimpe, and Kwok (1996), Yau and Hueng (2000), and Levy and Deshbakhsk (2003). However, the only relevant application of this methodology to the property market is found in the work of Wang (2000).

2. It has been proved by Cochrane (1988) that the limit of is indeed the innovation variance of the random walk component.

3. Akaike information criterion suggests 11 lags for the VAR.

4. The typical tests of the constancy of the parameters in the cointegration test is to analyse the sample path of the elements of the estimated impact matrix Π = αβ′. However, Hansen and Johansen (1999) argued that a fluctuation test of Π could be biased due to the distribution of the test (such as, Quintos 1995) and the asymptotic distributions of αand β (such as Quintos 1997). Therefore, in their test, instead of the test of the elements of Π, the time paths of the estimated eigenvalues, , (i = 1, ..., r) is evaluated. The changes in α and β is reflected in the estimated eigenvalues.

5. We also construct a rural LLA test. In order to constraint to six regions, we combine Linköping LLA and Örebro LLA. We only find price in Helsingbore LLA is significant on rural LLA. Rural and urban areas have different economic structure, limited transportation for work, and different housing market. Therefore, their connection would be expected lower and their causality channel might be different from that between urbans, which need further study.